Ola-7B: benchmark results
Provider: Tencent. Released 2025-02-07. Access: Open.
Aggregate Rating 1557 ± 21, rank #1595 of 10916 rated models, from 243 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AVI-Bench - Visual-Referenced Audio Hallucination | 83.2 | Task score (%) on visual-referenced audio hallucination (250 | 100 |
| MTAVG-Bench - Perceptual Speech Quality | 55.96 | Score (%) on the questions diagnosing acoustic integrity of | 100 |
| OpenVLM MMBench V1.0 CN - AR | 90.6 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.0 EN - AR | 91.3 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.1 CN - AR | 93.1 | Accuracy (%) | 99.8 |
| OpenVLM MM-Vet - Knowledge | 82.6 | Score | 99.6 |
| OpenVLM MM-Vet - Language Generation | 81 | Score | 99.3 |
| OpenVLM MMStar - Instance Reasoning | 85.6 | Accuracy (%) | 99.3 |
| OpenVLM MM-Vet - Recognition | 84 | Score | 98.9 |
| OpenVLM MMBench V1.1 CN - Function Reasoning | 96.7 | Accuracy (%) | 98.9 |
| OpenVLM MMBench V1.1 CN - Physical Property Reasoning | 82.3 | Accuracy (%) | 98.9 |
| OpenVLM MMStar - Coarse Perception | 81.2 | Accuracy (%) | 98.9 |
Interactive version: theaggregate.ai/model?slug=ola-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.